Superblocks Reveals AI Coding Secrets with 19 System Prompts
Brad Menezes, CEO of Superblocks, made waves in the AI community by publicly sharing system prompts from 19 prominent AI coding tools. This rare glimpse into the typically guarded world of prompt engineering came alongside the launch of Clark, Superblocks' enterprise AI agent product.
The Hidden Language of AI These system prompts—from tools like Windsurf, Manus, and Cursor—reveal how companies train models to specialize in specific domains. "Each company has completely different system prompts for the same large model," Menezes told TechCrunch. He describes these prompts as a "public course in prompt engineering," exposing both model behavior and startup strategies.
Beyond the Prompt: The Real Magic While revealing these examples, Menezes emphasizes that system prompts represent just 20% of the equation. The remaining 80% lies in sophisticated prompt engineering—how instructions are structured, how errors are corrected, and how results are optimized. "Every single thing you teach AI must be clearly expressed in human-level language," he explains.
The Superblocks team categorizes prompt engineering into three components:
- Role Prompt: Defines the model's identity and goals
- Situation Prompt: Sets task boundaries and rules
- Tool Invocation Prompt: Instructs the model on triggering external functions
Clark: AI for Non-Technical Teams Fresh off a $23 million Series A round (totaling $60 million), Superblocks positions Clark as an "enterprise internal AI agent" for sales and operations teams. Unlike developer-focused tools like Devin or Replit, Clark enables non-technical users to build custom internal systems—from CRM assistants to performance dashboards.
Menezes tested this concept internally by having operations teams build their own tools while engineers focused on product development. The experiment proved successful, demonstrating Clark's potential to democratize enterprise tool creation.
The Future of Prompt Engineering As AI tools become more standardized, system prompts and their accompanying mechanisms may emerge as key differentiators. "The smartest startup ideas are hidden behind complex system prompts," Menezes observes. These prompts represent not just technical instructions but the unique DNA of each AI product.
With prompt engineering moving from backstage to center stage, companies mastering this craft could define the next generation of AI tools.
Key Points
- Superblocks disclosed system prompts from 19 leading AI coding tools
- System prompts account for only 20% of a product's secret sauce
- Clark targets non-technical users for enterprise tool creation
- The company recently raised $23M in Series A funding
- Prompt engineering is becoming a crucial competitive advantage
